Learn how AI can enhance security measures by detecting malicious assets, and safeguarding against vulnerabilities. Stay secure with advanced technology.
The internet is huge and is expanding every day. While this has many positives for businesses, managing the potential risks in this environment can be daunting. The reality is that across the internet brands are susceptible to everything from brand abuse to phishing websites. So, how is it possible that organisations can keep track of the web-based assets that belong to them, and at the same time distinguish them from something that may only look like it’s theirs? Finding and verifying all of a company’s web assets across the entire internet is a massive undertaking. You essentially need to filter the whole internet and try to pick out what is relevant, and then set about detecting the risks – or even potential risks – within what you have found.
This isn’t a process that can be managed manually. The staff-hours alone would make this hugely prohibitive, and that’s without taking into account the potential margin for error. Instead, it requires a different approach, one based around automation. At Darktrace, my team and I work on exactly those kind of solutions. We’ve developed our own algorithms to define what distinguishes a client’s brand-owned site from everything else on the internet. We refer to that as a company’s Brand DNA. These special characteristics help us to predict and identify where any brand-related assets are across the entire internet, and how they should be investigated further. The concept of an organisation’s Brand DNA breaks down into two areas: what is unique by design and what is uniqueby comparison.
Unique by Design
All brands have different design elements that help set them apart from other brands. This can be everything from their name and logo, to the different fonts and colours they use in all their communications and websites. By ingesting this data into our algorithms we are able to scan the internet for any web-based assets that may be relevant to a company, based on these key brand elements. While the elements of ‘unique by design’ are relatively easily understandable by humans, no organisations want to have their time taken up manually searching through millions of images every day in order to help them locate the web properties that might belong to their organisation.
Unique by Comparison
Conversely, ‘unique by comparison’ focuses more on the elements that a human would probably not be able to figure out by themselves. As we suggest potential new domains to customers, we build up a pool of assets. Automation allows us to find patterns in these assets that might not be immediately obvious to humans, such as elements of metadata,nameserver details, or even where a website is hosted. Although unique by design is more about what you actually see on a website and unique by comparison focuses on the back-end, in reality there are overlaps and the two things feed into each other.
As a very basic example: if a domain is hosted on nameserver where other company assets are hosted, AND the company logo is on the page, then the chances are this domain is a company-owned asset. In effect, the two approaches strengthen each other. By analysing all these details together, our algorithms can increasingly accurately score how likely an asset is to be owned by a company. I should add here that unique by comparison is based on comparing a lot of features at once, so it is often not as clear cut as the above example.
Combining Humans and AI
Ultimately, automating the process in this way helps to create a minimal-touch process for companies. The algorithms do all the filtering, enabling the creation of a much-reduced list of assets for the company to look through. Basically, we’re able to break down that list to avery small percentage of the internet that they actually need to look at and then analyse the risk those assets poseto the organisation.
We also use something which we internally label “AI2”(artificial intelligence with analyst interaction). This essentially means we’re adding a human layer to the automated process, for both input and output checks. While the algorithms do all the heavy lifting and aid scalability, the human element allows us to finetune or dive deeper into certain automated findings.
Detecting Malicious Assets
While the algorithms are principally focused on establishing a brand’s attack surface, a useful byproduct is that they can also locate malicious assets, such as potential phishing sites. For example, if something looks like it belongs to the customer, but doesn’t actually belong inside their directdigital infrastructure, then clearly there is an increased likelihood that it is either a brand abuse or phishing site.
On top of this, as part of our search process we can also automatically create combinations of possible URLs that cover common search errors such as typos or “fat finger”errors within brand names, and then hunt for those – clearly the likelihood of these URLs being rogue sites is greatly enhanced.
The Clearest View of the Attack Surface
By combining all these elements, companies are able to get the most complete view of their potential attack surface.And with the use of enhanced automation techniques they can do so with minimum effort. From this position companies are able to easily and quickly home in on the genuine items, and the areas that pose them the most risk. They can then use the resulting list to form the foundations from which they can apply the rest of their security strategy.
Like this and want more?
Receive the latest blog in your inbox
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
NEWSLETTER
Like this and want more?
Stay up to date on the latest industry news and insights.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Disarming the WarmCookie Backdoor: Darktrace’s Oven-Ready Solution
26
Jul 2024
What is WarmCookie malware?
WarmCookie, also known as BadSpace [2], is a two-stage backdoor tool that provides functionality for threat actors to retrieve victim information and launch additional payloads. The malware is primarily distributed via phishing campaigns according to multiple open-source intelligence (OSINT) providers.
Backdoor malware: A backdoor tool is a piece of software used by attackers to gain and maintain unauthorized access to a system. It bypasses standard authentication and security mechanisms, allowing the attacker to control the system remotely.
Two-stage backdoor malware: This means the backdoor operates in two distinct phases:
1. Initial Stage: The first stage involves the initial infection and establishment of a foothold within the victim's system. This stage is often designed to be small and stealthy to avoid detection.
2. Secondary Stage: Once the initial stage has successfully compromised the system, it retrieves or activates the second stage payload. This stage provides more advanced functionalities for the attacker, such as extensive data exfiltration, deeper system control, or the deployment of additional malicious payloads.
How does WarmCookie malware work?
Reported attack patterns include emails attempting to impersonate recruitment firms such as PageGroup, Michael Page, and Hays. These emails likely represented social engineering tactics, with attackers attempting to manipulate jobseekers into engaging with the emails and following malicious links embedded within [3].
This backdoor tool also adopts stealth and evasion tactics to avoid the detection of traditional security tools. Reported evasion tactics included custom string decryption algorithms, as well as dynamic API loading to prevent researchers from analyzing and identifying the core functionalities of WarmCookie [1].
Before this backdoor makes an outbound network request, it is known to capture details from the target machine, which can be used for fingerprinting and identification [1], this includes:
- Computer name
- Username
- DNS domain of the machine
- Volume serial number
WarmCookie samples investigated by external researchers were observed communicating communicated over HTTP to a hardcoded IP address using a combination of RC4 and Base64 to protect its network traffic [1]. Ultimately, threat actors could use this backdoor to deploy further malicious payloads on targeted networks, such as ransomware.
Darktrace Coverage of WarmCookie
Between April and June 2024, Darktrace’s Threat Research team investigated suspicious activity across multiple customer networks indicating that threat actors were utilizing the WarmCookie backdoor tool. Observed cases across customer environments all included the download of unusual executable (.exe) files and suspicious outbound connectivity.
Affected devices were all observed making external HTTP requests to the German-based external IP, 185.49.69[.]41, and the URI, /data/2849d40ade47af8edfd4e08352dd2cc8.
Less than a minute later, the same device was observed making HTTP requests to the rare external IP address: 185.49.69[.]41, which had never previously been observed on the network, for the URI /data/b834116823f01aeceed215e592dfcba7. The device then proceeded to download masqueraded executable file from this endpoint. Darktrace recognized that these connections to an unknown endpoint, coupled with the download of a masqueraded file, likely represented malicious activity.
Following this download, the device began beaconing back to the same IP, 185.49.69[.]41, with a large number of external connections observed over port 80. This beaconing related behavior could further indicate malicious software communicating with command-and-control (C2) servers.
Darktrace’s model alert coverage included the following details:
[Model Alert: Device / Unusual BITS Activity]
- Associated device type: desktop
- Time of alert: 2024-04-23T14:10:23 UTC
- ASN: AS28753 Leaseweb Deutschland GmbH
- User agent: Microsoft BITS/7.8
[Model Alert: Anomalous File / EXE from Rare External Location]
[Model Alert: Compromise / Beaconing Activity To External Rare]
- Associated device type: desktop
- Time of alert: 2024-04-23T14:15:24 UTC
- Destination IP: 185.49.69[.]41
- Destination port: 80
- Protocol: TCP
- Application protocol: HTTP
- ASN: AS28753 Leaseweb Deutschland GmbH
- User agent: Mozilla / 4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1;.NET CLR 1.0.3705)
Between May 7 and June 4, Darktrace identified a wide range of suspicious external connectivity on another customer’s environment. Darktrace’s Threat Research team further investigated this activity and assessed it was likely indicative of WarmCookie exploitation on customer devices.
Similar to the initial use case, BITS activity was observed on affected devices, which is utilized to download WarmCookie [1]. This initial behavior was observed with the device after triggering the model: Device / Unusual BITS Activity on May 7.
Just moments later, the same device was observed making HTTP requests to the aforementioned German IP address, 185.49.69[.]41 using the same URI /data/2849d40ade47af8edfd4e08352dd2cc8, before downloading a suspicious executable file.
Just like the first use case, this device followed up this suspicious download with a series of beaconing connections to 185.49.69[.]41, again with a large number of connections via port 80.
Similar outgoing connections to 185.49.69[.]41 and model alerts were observed on additional devices during the same timeframe, indicating that numerous customer devices had been compromised.
Darktrace’s model alert coverage included the following details:
[Model Alert: Device / Unusual BITS Activity]
- Associated device type: desktop
- Time of alert: 2024-05-07T09:03:23 UTC
- ASN: AS28753 Leaseweb Deutschland GmbH
- User agent: Microsoft BITS/7.8
[Model Alert: Anomalous File / EXE from Rare External Location]
[Model Alert: Anomalous Connection / New User Agent to IP Without Hostname]
- Associated device type: desktop
- Time of alert: 2024-05-07T09:04:14 UTC
- Destination IP: 185.49.69[.]41
- Application protocol: HTTP
- URI: /data/2849d40ade47af8edfd4e08352dd2cc8
- User agent: Microsoft BITS/7.8
[Model Alert: Compromise / HTTP Beaconing to New Endpoint]
- Associated device type: desktop
- Time of alert: 2024-05-07T09:08:47 UTC
- Destination IP: 185.49.69[.]41
- Protocol: TCP
- Application protocol: HTTP
- ASN: AS28753 Leaseweb Deutschland GmbH
- URI: /
- User agent: Mozilla / 4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1;.NET CLR 1.0.3705) \
Figure 1: Cyber AI Analyst Coverage Details around the external destination, ‘185.49.69[.]41’.
Figure 2: External Sites Summary verifying the geographical location of the external IP, 185.49.69[.]41’.
Fortunately, this particular customer was subscribed to Darktrace’s Proactive Threat Notification (PTN) service and the Darktrace Security Operation Center (SOC) promptly investigated the activity and alerted the customer. This allowed their security team to address the activity and begin their own remediation process.
In this instance, Darktrace’s Autonomous Response capability was configured in Human Confirmation mode, meaning any mitigative actions required manual application by the customer’s security team.
Despite this, Darktrace recommended two actions to contain the activity: blocking connections to the suspicious IP address 185.49.69[.]41 and any IP addresses ending with '69[.]41', as well as the ‘Enforce Pattern of Life’ action. By enforcing a pattern of life, Darktrace can restrict a device (or devices) to its learned behavior, allowing it to continue regular business activities uninterrupted while blocking any deviations from expected activity.
Figure 3: Actions suggested by Darktrace to contain the emerging activity, including blocking connections to the suspicious endpoint and restricting the device to its ‘pattern of life’.
Conclusion
Backdoor tools like WarmCookie enable threat actors to gather and leverage information from target systems to deploy additional malicious payloads, escalating their cyber attacks. Given that WarmCookie’s primary distribution method seems to be through phishing campaigns masquerading as trusted recruitments firms, it has the potential to affect a large number of organziations.
In the face of such threats, Darktrace’s behavioral analysis provides organizations with full visibility over anomalous activity on their digital estates, regardless of whether the threat bypasses by human security teams or email security tools. While threat actors seemingly managed to evade customers’ native email security and gain access to their networks in these cases, Darktrace identified the suspicious behavior associated with WarmCookie and swiftly notified customer security teams.
Had Darktrace’s Autonomous Response capability been fully enabled in these cases, it could have blocked any suspicious connections and subsequent activity in real-time, without the need of human intervention, effectively containing the attacks in the first instance.
Credit to Justin Torres, Cyber Security Analyst and Dylan Hinz, Senior Cyber Security Analyst
Appendices
Darktrace Model Detections
- Anomalous File / EXE from Rare External Location
- Anomalous File / Masqueraded File Transfer
- Compromise / Beacon to Young Endpoint
- Compromise / Beaconing Activity To External Rare
- Compromise / HTTP Beaconing to New Endpoint
- Compromise / HTTP Beaconing to Rare Destination
- Compromise / High Volume of Connections with Beacon Score
- Compromise / Large Number of Suspicious Successful Connections
- Compromise / Quick and Regular Windows HTTP Beaconing
- Compromise / SSL or HTTP Beacon
- Compromise / Slow Beaconing Activity To External Rare
- Compromise / Sustained SSL or HTTP Increase
- Compromise / Sustained TCP Beaconing Activity To Rare Endpoint
- Anomalous Connection / Multiple Failed Connections to Rare Endpoint
- Anomalous Connection / New User Agent to IP Without Hostname
- Compromise / Sustained SSL or HTTP Increase
AI Analyst Incident Coverage:
- Unusual Repeated Connections
- Possible SSL Command and Control to Multiple Endpoints
The State of AI in Cybersecurity: Understanding AI Technologies
24
Jul 2024
About the State of AI Cybersecurity Report
Darktrace surveyed 1,800 CISOs, security leaders, administrators, and practitioners from industries around the globe. Our research was conducted to understand how the adoption of new AI-powered offensive and defensive cybersecurity technologies are being managed by organizations.
How familiar are security professionals with supervised machine learning
Just 31% of security professionals report that they are “very familiar” with supervised machine learning.
Many participants admitted unfamiliarity with various AI types. Less than one-third felt "very familiar" with the technologies surveyed: only 31% with supervised machine learning and 28% with natural language processing (NLP).
Most participants were "somewhat" familiar, ranging from 46% for supervised machine learning to 36% for generative adversarial networks (GANs). Executives and those in larger organizations reported the highest familiarity.
Combining "very" and "somewhat" familiar responses, 77% had familiarity with supervised machine learning, 74% generative AI, and 73% NLP. With generative AI getting so much media attention, and NLP being the broader area of AI that encompasses generative AI, these results may indicate that stakeholders are understanding the topic on the basis of buzz, not hands-on work with the technologies.
If defenders hope to get ahead of attackers, they will need to go beyond supervised learning algorithms trained on known attack patterns and generative AI. Instead, they’ll need to adopt a comprehensive toolkit comprised of multiple, varied AI approaches—including unsupervised algorithms that continuously learn from an organization’s specific data rather than relying on big data generalizations.
Different types of AI
Different types of AI have different strengths and use cases in cyber security. It’s important to choose the right technique for what you’re trying to achieve.
Supervised machine learning: Applied more often than any other type of AI in cyber security. Trained on human attack patterns and historical threat intelligence.
Large language models (LLMs): Applies deep learning models trained on extremely large data sets to understand, summarize, and generate new content. Used in generative AI tools.
Natural language processing (NLP): Applies computational techniques to process and understand human language.
Unsupervised machine learning: Continuously learns from raw, unstructured data to identify deviations that represent true anomalies.
What impact will generative AI have on the cybersecurity field?
More than half of security professionals (57%) believe that generative AI will have a bigger impact on their field over the next few years than other types of AI.
Figure 1: Chart from Darktrace's State of AI in Cybersecurity Report
Security stakeholders are highly aware of generative AI and LLMs, viewing them as pivotal to the field's future. Generative AI excels at abstracting information, automating tasks, and facilitating human-computer interaction. However, LLMs can "hallucinate" due to training data errors and are vulnerable to prompt injection attacks. Despite improvements in securing LLMs, the best cyber defenses use a mix of AI types for enhanced accuracy and capability.
AI education is crucial as industry expectations for generative AI grow. Leaders and practitioners need to understand where and how to use AI while managing risks. As they learn more, there will be a shift from generative AI to broader AI applications.
Do security professionals fully understand the different types of AI in security products?
Only 26% of security professionals report a full understanding of the different types of AI in use within security products.
Confusion is prevalent in today’s marketplace. Our survey found that only 26% of respondents fully understand the AI types in their security stack, while 31% are unsure or confused by vendor claims. Nearly 65% believe generative AI is mainly used in cybersecurity, though it’s only useful for identifying phishing emails. This highlights a gap between user expectations and vendor delivery, with too much focus on generative AI.
Key findings include:
Executives and managers report higher understanding than practitioners.
Larger organizations have better understanding due to greater specialization.
As AI evolves, vendors are rapidly introducing new solutions faster than practitioners can learn to use them. There's a strong need for greater vendor transparency and more education for users to maximize the technology's value.
To help ease confusion around AI technologies in cybersecurity, Darktrace has released the CISO’s Guide to Cyber AI. A comprehensive white paper that categorizes the different applications of AI in cybersecurity. Download the White Paper here.
Do security professionals believe generative AI alone is enough to stop zero-day threats?
No! 86% of survey participants believe generative AI alone is NOT enough to stop zero-day threats
This consensus spans all geographies, organization sizes, and roles, though executives are slightly less likely to agree. Asia-Pacific participants agree more, while U.S. participants agree less.
Despite expecting generative AI to have the most impact, respondents recognize its limited security use cases and its need to work alongside other AI types. This highlights the necessity for vendor transparency and varied AI approaches for effective security across threat prevention, detection, and response.
Stakeholders must understand how AI solutions work to ensure they offer advanced, rather than outdated, threat detection methods. The survey shows awareness that old methods are insufficient.
![Cyber AI Analyst Coverage Details around the external destination, ‘185.49.69[.]41’.](https://cdn.prod.website-files.com/626ff4d25aca2edf4325ff97/66a3e9519710d78dcbea06b5_66a3de28f7ac7a01b4b94b45_Screenshot%25202024-07-26%2520at%252010.33.48%2520AM.png)
![External Sites Summary verifying the geographical location of the external IP, 185.49.69[.]41’.](https://cdn.prod.website-files.com/626ff4d25aca2edf4325ff97/66a3e9519710d78dcbea06ad_66a3de6808c6bac429c4520e_Screenshot%25202024-07-26%2520at%252010.34.08%2520AM.png)
